sharp/docs/performance.md
Lovell Fuller ca102ebd6c Docs: fix perf result copypasta
An ARM64 value was incorrectly using an AMD64 result
2023-03-28 12:08:07 +01:00

112 lines
4.7 KiB
Markdown

# Performance
A test to benchmark the performance of this module relative to alternatives.
Greater libvips performance can be expected with caching enabled (default)
and using 8+ core machines, especially those with larger L1/L2 CPU caches.
The I/O limits of the relevant (de)compression library will generally determine maximum throughput.
## Contenders
* [jimp](https://www.npmjs.com/package/jimp) v0.22.7 - Image processing in pure JavaScript.
* [imagemagick](https://www.npmjs.com/package/imagemagick) v0.1.3 - Supports filesystem only and "*has been unmaintained for a long time*".
* [gm](https://www.npmjs.com/package/gm) v1.25.0 - Fully featured wrapper around GraphicsMagick's `gm` command line utility.
* [@squoosh/lib](https://www.npmjs.com/package/@squoosh/lib) v0.4.0 - Image libraries transpiled to WebAssembly, includes GPLv3 code, but "*Project no longer maintained*".
* [@squoosh/cli](https://www.npmjs.com/package/@squoosh/cli) v0.7.3 - Command line wrapper around `@squoosh/lib`, avoids GPLv3 by spawning process, but "*Project no longer maintained*".
* sharp v0.32.0 / libvips v8.14.2 - Caching within libvips disabled to ensure a fair comparison.
## Environment
### AMD64
* AWS EC2 us-east-2 [c6a.xlarge](https://aws.amazon.com/ec2/instance-types/c6a/) (4x AMD EPYC 7R13)
* Ubuntu 22.04 20230303 (ami-0122295b0eb922138)
* Node.js 16.19.1
### ARM64
* AWS EC2 us-east-2 [c7g.xlarge](https://aws.amazon.com/ec2/instance-types/c7g/) (4x ARM Graviton3)
* Ubuntu 22.04 20230303 (ami-0af198159897e7a29)
* Node.js 16.19.1
## Task: JPEG
Decompress a 2725x2225 JPEG image,
resize to 720x588 using Lanczos 3 resampling (where available),
then compress to JPEG at a "quality" setting of 80.
Note: jimp does not support Lanczos 3, bicubic resampling used instead.
#### Results: JPEG (AMD64)
| Module | Input | Output | Ops/sec | Speed-up |
| :----------------- | :----- | :----- | ------: | -------: |
| jimp | buffer | buffer | 0.84 | 1.0 |
| squoosh-cli | file | file | 1.07 | 1.3 |
| squoosh-lib | buffer | buffer | 1.82 | 2.2 |
| gm | buffer | buffer | 8.41 | 10.0 |
| gm | file | file | 8.45 | 10.0 |
| imagemagick | file | file | 8.77 | 10.4 |
| sharp | stream | stream | 36.36 | 43.3 |
| sharp | file | file | 38.67 | 46.0 |
| sharp | buffer | buffer | 39.44 | 47.0 |
#### Results: JPEG (ARM64)
| Module | Input | Output | Ops/sec | Speed-up |
| :----------------- | :----- | :----- | ------: | -------: |
| jimp | buffer | buffer | 1.02 | 1.0 |
| squoosh-cli | file | file | 1.11 | 1.1 |
| squoosh-lib | buffer | buffer | 2.08 | 2.0 |
| gm | buffer | buffer | 8.80 | 8.6 |
| gm | file | file | 10.05 | 9.9 |
| imagemagick | file | file | 10.28 | 10.1 |
| sharp | stream | stream | 26.87 | 26.3 |
| sharp | file | file | 27.88 | 27.3 |
| sharp | buffer | buffer | 28.40 | 27.8 |
## Task: PNG
Decompress a 2048x1536 RGBA PNG image,
premultiply the alpha channel,
resize to 720x540 using Lanczos 3 resampling (where available),
unpremultiply then compress as PNG with a "default" zlib compression level of 6
and without adaptive filtering.
Note: jimp does not support premultiply/unpremultiply.
### Results: PNG (AMD64)
| Module | Input | Output | Ops/sec | Speed-up |
| :----------------- | :----- | :----- | ------: | -------: |
| squoosh-cli | file | file | 0.40 | 1.0 |
| squoosh-lib | buffer | buffer | 0.47 | 1.2 |
| gm | file | file | 6.47 | 16.2 |
| jimp | buffer | buffer | 6.60 | 16.5 |
| imagemagick | file | file | 7.08 | 17.7 |
| sharp | file | file | 17.80 | 44.5 |
| sharp | buffer | buffer | 18.02 | 45.0 |
### Results: PNG (ARM64)
| Module | Input | Output | Ops/sec | Speed-up |
| :----------------- | :----- | :----- | ------: | -------: |
| squoosh-cli | file | file | 0.40 | 1.0 |
| squoosh-lib | buffer | buffer | 0.48 | 1.2 |
| gm | file | file | 7.20 | 18.0 |
| jimp | buffer | buffer | 7.62 | 19.1 |
| imagemagick | file | file | 7.96 | 19.9 |
| sharp | file | file | 12.97 | 32.4 |
| sharp | buffer | buffer | 13.12 | 32.8 |
## Running the benchmark test
Requires Docker.
```sh
git clone https://github.com/lovell/sharp.git
cd sharp/test/bench
./run-with-docker.sh
```